All Intel processors affected by memory leaking vulnerability

[Koenig168]

Anyone on Sandy Bridge architecture and experienced any performance hit?

I on i5-2500K and I don't really see any performance hit.

IIRC have to combine the Windows update and new BIOS to see any meaningful performance hit. I don't think the BIOS updates have rolled out to Sandy Bridge MBs yet.

 

[Encrypted11]

Anyone on Sandy Bridge architecture and experienced any performance hit?

I on i5-2500K and I don't really see any performance hit.

If you properly applied the patch, you should see the following indicating meltdown patch is active. Spectre needs microcode.

Refer to this for the meaning..

For Windows:
https://support.microsoft.com/en-sg...-of-get-speculationcontrolsettings-powershell

Getting that powershell moduleL https://www.powershellgallery.com/packages/SpeculationControl/1.0.4

Would doubt its that negligible on Sandybridge in a wider variety of tasks. Ageing pcie 2.0 bus that are easier to saturate, no native USB 3.0 etc.

 

[Encrypted11]

i didn't do any proper testing, just daily copying video files over to my other old pc x3 720be with realtek lan - same 111MB/s as before - any hit won't be shown in my setup

You enable the 4th core will add more throughput?

Can add IO FUD or not collected days ago?
HDD impact will probably be lower in percentage terms but probably more perceptible?

 

[Ferolare]

IIRC have to combine the Windows update and new BIOS to see any meaningful performance hit. I don't think the BIOS updates have rolled out to Sandy Bridge MBs yet.

Probably safe to assume no update to preSkylake bah

 

[AnTiLooP]

Here's an update from me and also an observation:

based on my hardware, if I were to install only the OS side patches, windows for some reason appears extremely sluggish and I was able to reproduce the scenario where it would load past UEFI and into windows but hangs up with a mouse pointer on screen, nothing else. the HDD LED continues to blink continously during this time. Only hitting the reset button can i reboot and boot succesfully into Windows.

after installing the latest bios for my board, the symptoms went away completely.

I'm not sure if this is something worth to mention but for you folks with similar issues please note the above.

 

[12122012]

after update, windows feel the same only noticed boot time from 9+sec. to 11+sec. all SSDs DiskMark around the same too... no idea if there going to be any new update bios for my old mb... sigh

 

[hlots123]

You enable the 4th core will add more throughput?

Can add IO FUD or not collected days ago?
HDD impact will probably be lower in percentage terms but probably more perceptible?

i rem doing the unlock but can't rem the outcome now - more likely mine no 4th core iic

i forgot i'm copying to an old hdd, so 111MB/s could be old hdd limit
1000/8=125MB/s is quite close i totally didn't think of trying out ssd to ssd...

 

[elmariachi]

Looks like Icelake may not come in 2018 either.

http://host.madison.com/business/in...cle_fcf70bbe-0947-5062-b4da-2e882a3ddae6.html

 

[haylui]

You enable the 4th core will add more throughput?

Can add IO FUD or not collected days ago?
HDD impact will probably be lower in percentage terms but probably more perceptible?

Can see it from cinebench...lol

 

[cscs3]

And the fixes for it are expected to introduce a performance penalty of up to 50%.

As reported by Phoronix on the Linux side of things:



Of course, with Linux being open source, fixes have already been committed and mainlined into the upcoming kernel 4.15. The article claims a performance penalty of up to 30%, but Phoronix's benchmark on the kernel shows more than 50% I/O penalty on a CoffeeLake:

It will be interesting to see when Windows and macOS will receive their respective kernel patches and the extent of the performance penalty introduced there.

Article source: https://www.phoronix.com/vr.php?view=25767

Also interesting to note: the kernel developers are treating all x86 processors as being affected:

Your title is very misleading.
Your message say : the kernel developers are treating all x86 processors as being affected.

AMD processor is not x86?

 

[Rock-kun]

Your title is very misleading.
Your message say : the kernel developers are treating all x86 processors as being affected.

AMD processor is not x86?

First of all, that post is almost 2 weeks old?

Secondly, Meltdown is known to be Intel-specifc so I'm not wrong.

And the initial patch to the Linux kernel was to label all x86 processors (including AMD) as being affected even though AMD is not (it raised a little hooha in LKML, but that's besides the point).

Blanket labeling all x86 hardware as 'affected' != 'AMD is affected'. No one ever said anything about Meltdown on AMD. I can edit the kernel source and set a string claiming that even ARM, PPC and SPARC are affected by Meltdown, if I wanted to. Are you then going to tell the world that all their switches, routers, smartphones and supercomputers are affected by Meltdown?

Finally, that line of code in the Linux kernel was eventually modified to target only Intel hardware. So I'm still not wrong in any way.

 

[hlots123]

Probably safe to assume no update to preSkylake bah

i just saw beta bios for my asus b85 board, dunno if its meltdown/spectre fix but suspect likely so...
https://www.asus.com/sg/Motherboards/B85MG/HelpDesk_Download/

 

[Ferolare]

Asrock on the other hand...

Lucky linux has the microcode up since early January

 

[Rock-kun]

Asrock on the other hand...

Lucky linux has the microcode up since early January

That's because Linux lacks UEFI updating tools. Most board firmware updates are typically done from within Windows with a tool provided by the board vendors themselves.

In the absence of such tools, the Linux kernel has facilities in place to upload updated microcode to the processor on the fly as part of the boot process. I believe Windows must have something similar as well.

And I'm also 100% certain that the updates released by the board vendors are using the same microcode released for Linux. The difference is that your firmware updates are such that the updated microcode is uploaded to the processor as part of POST, while Linux does it as part of the boot process.

 

[Encrypted11]

Can see it from cinebench...lol

Where did you see the numbers or you made them up?

The 2nd stick adding to dual channel memory array in my current run vs. setup/diagnostic (1 stick + dual tower air sink) phase costs about +/-20 cinebench points.

I think most people here are trying to patch a security hole and weigh the performance impact than to jjww. People in the glass house shouldn't throw stones.

 

[hlots123]

[TECHSPOT] Patched Desktop PC: Meltdown & Spectre Benchmarked
^ i3-8100 b4 update, win update, win+bios update compared

 

[Encrypted11]

[TECHSPOT] Patched Desktop PC: Meltdown & Spectre Benchmarked
^ i3-8100 b4 update, win update, win+bios update compared

the stilt hevc decoding benchmark 1.61 lost about 1 second too

 

[Ferolare]

I think most people here are trying to patch a security hole and weigh the performance impact than to jjww. People in the glass house shouldn't throw stones.

And yet throw i must

*throw stone at asrock* when can i expect a patch at the hardware level instead of relying on bandaid?

*throw more stones at asrock* when can I return to Windows and upgrade my rig to be 4k hdr capable?



(from a performance pov, nope. Nothing to report as the usage pattern shouldn't see reduced perfomance at all: 1080p60 twitch stream is still retardedly taxing on my i5)

 

[Encrypted11]

And yet throw i must

*throw stone at asrock* when can i expect a patch at the hardware level instead of relying on bandaid?

*throw more stones at asrock* when can I return to Windows and upgrade my rig to be 4k hdr capable?



(from a performance pov, nope. Nothing to report as the usage pattern shouldn't see reduced perfomance at all: 1080p60 twitch stream is still retardedly taxing on my i5)

That's what we paid for buying non-#1 motherboard with even the slowest microcode update at the am3+ times is it?

But if you wanna self patch, the trick is indeed the win raid UBU.

Need to check if the bios it spits out is the exact kb size as original (~16MB due to 128Mb ROM chip), the downside though is if it nukes your BIOS in a single BIOS board you might be hunting for a 128Mb ROM chip and spi flash programmer on taobao.

But the mod itself takes less than a min.

 

[Ferolare]

Risk big, win big.

Anything to avoid paying through the nose for DDR4 memory kit.

I might do a quick x264 bench when I have the time to spare. After all, encoding something something via x264 doesn't take a long arse time to finish and it can still be somewhat relevant to an Intel haswell i5 cpu